xref: /linux/kernel/audit.c (revision cbafa54aa2ae23939846e150ad4ba98c784f6395)
1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /* audit.c -- Auditing support
3  * Gateway between the kernel (e.g., selinux) and the user-space audit daemon.
4  * System-call specific features have moved to auditsc.c
5  *
6  * Copyright 2003-2007 Red Hat Inc., Durham, North Carolina.
7  * All Rights Reserved.
8  *
9  * Written by Rickard E. (Rik) Faith <faith@redhat.com>
10  *
11  * Goals: 1) Integrate fully with Security Modules.
12  *	  2) Minimal run-time overhead:
13  *	     a) Minimal when syscall auditing is disabled (audit_enable=0).
14  *	     b) Small when syscall auditing is enabled and no audit record
15  *		is generated (defer as much work as possible to record
16  *		generation time):
17  *		i) context is allocated,
18  *		ii) names from getname are stored without a copy, and
19  *		iii) inode information stored from path_lookup.
20  *	  3) Ability to disable syscall auditing at boot time (audit=0).
21  *	  4) Usable by other parts of the kernel (if audit_log* is called,
22  *	     then a syscall record will be generated automatically for the
23  *	     current syscall).
24  *	  5) Netlink interface to user-space.
25  *	  6) Support low-overhead kernel-based filtering to minimize the
26  *	     information that must be passed to user-space.
27  *
28  * Audit userspace, documentation, tests, and bug/issue trackers:
29  * 	https://github.com/linux-audit
30  */
31 
32 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
33 
34 #include <linux/file.h>
35 #include <linux/init.h>
36 #include <linux/types.h>
37 #include <linux/atomic.h>
38 #include <linux/mm.h>
39 #include <linux/export.h>
40 #include <linux/slab.h>
41 #include <linux/err.h>
42 #include <linux/kthread.h>
43 #include <linux/kernel.h>
44 #include <linux/syscalls.h>
45 #include <linux/spinlock.h>
46 #include <linux/rcupdate.h>
47 #include <linux/mutex.h>
48 #include <linux/gfp.h>
49 #include <linux/pid.h>
50 
51 #include <linux/audit.h>
52 
53 #include <net/sock.h>
54 #include <net/netlink.h>
55 #include <linux/skbuff.h>
56 #ifdef CONFIG_SECURITY
57 #include <linux/security.h>
58 #endif
59 #include <linux/freezer.h>
60 #include <linux/pid_namespace.h>
61 #include <net/netns/generic.h>
62 
63 #include "audit.h"
64 
65 /* No auditing will take place until audit_initialized == AUDIT_INITIALIZED.
66  * (Initialization happens after skb_init is called.) */
67 #define AUDIT_DISABLED		-1
68 #define AUDIT_UNINITIALIZED	0
69 #define AUDIT_INITIALIZED	1
70 static int	audit_initialized = AUDIT_UNINITIALIZED;
71 
72 u32		audit_enabled = AUDIT_OFF;
73 bool		audit_ever_enabled = !!AUDIT_OFF;
74 
75 EXPORT_SYMBOL_GPL(audit_enabled);
76 
77 /* Default state when kernel boots without any parameters. */
78 static u32	audit_default = AUDIT_OFF;
79 
80 /* If auditing cannot proceed, audit_failure selects what happens. */
81 static u32	audit_failure = AUDIT_FAIL_PRINTK;
82 
83 /* private audit network namespace index */
84 static unsigned int audit_net_id;
85 
86 /**
87  * struct audit_net - audit private network namespace data
88  * @sk: communication socket
89  */
90 struct audit_net {
91 	struct sock *sk;
92 };
93 
94 /**
95  * struct auditd_connection - kernel/auditd connection state
96  * @pid: auditd PID
97  * @portid: netlink portid
98  * @net: the associated network namespace
99  * @rcu: RCU head
100  *
101  * Description:
102  * This struct is RCU protected; you must either hold the RCU lock for reading
103  * or the associated spinlock for writing.
104  */
105 struct auditd_connection {
106 	struct pid *pid;
107 	u32 portid;
108 	struct net *net;
109 	struct rcu_head rcu;
110 };
111 static struct auditd_connection __rcu *auditd_conn;
112 static DEFINE_SPINLOCK(auditd_conn_lock);
113 
114 /* If audit_rate_limit is non-zero, limit the rate of sending audit records
115  * to that number per second.  This prevents DoS attacks, but results in
116  * audit records being dropped. */
117 static u32	audit_rate_limit;
118 
119 /* Number of outstanding audit_buffers allowed.
120  * When set to zero, this means unlimited. */
121 static u32	audit_backlog_limit = 64;
122 #define AUDIT_BACKLOG_WAIT_TIME (60 * HZ)
123 static u32	audit_backlog_wait_time = AUDIT_BACKLOG_WAIT_TIME;
124 
125 /* The identity of the user shutting down the audit system. */
126 static kuid_t		audit_sig_uid = INVALID_UID;
127 static pid_t		audit_sig_pid = -1;
128 static u32		audit_sig_sid;
129 
130 /* Records can be lost in several ways:
131    0) [suppressed in audit_alloc]
132    1) out of memory in audit_log_start [kmalloc of struct audit_buffer]
133    2) out of memory in audit_log_move [alloc_skb]
134    3) suppressed due to audit_rate_limit
135    4) suppressed due to audit_backlog_limit
136 */
137 static atomic_t	audit_lost = ATOMIC_INIT(0);
138 
139 /* Monotonically increasing sum of time the kernel has spent
140  * waiting while the backlog limit is exceeded.
141  */
142 static atomic_t audit_backlog_wait_time_actual = ATOMIC_INIT(0);
143 
144 /* Hash for inode-based rules */
145 struct list_head audit_inode_hash[AUDIT_INODE_BUCKETS];
146 
147 static struct kmem_cache *audit_buffer_cache;
148 
149 /* queue msgs to send via kauditd_task */
150 static struct sk_buff_head audit_queue;
151 /* queue msgs due to temporary unicast send problems */
152 static struct sk_buff_head audit_retry_queue;
153 /* queue msgs waiting for new auditd connection */
154 static struct sk_buff_head audit_hold_queue;
155 
156 /* queue servicing thread */
157 static struct task_struct *kauditd_task;
158 static DECLARE_WAIT_QUEUE_HEAD(kauditd_wait);
159 
160 /* waitqueue for callers who are blocked on the audit backlog */
161 static DECLARE_WAIT_QUEUE_HEAD(audit_backlog_wait);
162 
163 static struct audit_features af = {.vers = AUDIT_FEATURE_VERSION,
164 				   .mask = -1,
165 				   .features = 0,
166 				   .lock = 0,};
167 
168 static char *audit_feature_names[2] = {
169 	"only_unset_loginuid",
170 	"loginuid_immutable",
171 };
172 
173 /**
174  * struct audit_ctl_mutex - serialize requests from userspace
175  * @lock: the mutex used for locking
176  * @owner: the task which owns the lock
177  *
178  * Description:
179  * This is the lock struct used to ensure we only process userspace requests
180  * in an orderly fashion.  We can't simply use a mutex/lock here because we
181  * need to track lock ownership so we don't end up blocking the lock owner in
182  * audit_log_start() or similar.
183  */
184 static struct audit_ctl_mutex {
185 	struct mutex lock;
186 	void *owner;
187 } audit_cmd_mutex;
188 
189 /* AUDIT_BUFSIZ is the size of the temporary buffer used for formatting
190  * audit records.  Since printk uses a 1024 byte buffer, this buffer
191  * should be at least that large. */
192 #define AUDIT_BUFSIZ 1024
193 
194 /* The audit_buffer is used when formatting an audit record.  The caller
195  * locks briefly to get the record off the freelist or to allocate the
196  * buffer, and locks briefly to send the buffer to the netlink layer or
197  * to place it on a transmit queue.  Multiple audit_buffers can be in
198  * use simultaneously. */
199 struct audit_buffer {
200 	struct sk_buff       *skb;	/* formatted skb ready to send */
201 	struct audit_context *ctx;	/* NULL or associated context */
202 	gfp_t		     gfp_mask;
203 };
204 
205 struct audit_reply {
206 	__u32 portid;
207 	struct net *net;
208 	struct sk_buff *skb;
209 };
210 
211 /**
212  * auditd_test_task - Check to see if a given task is an audit daemon
213  * @task: the task to check
214  *
215  * Description:
216  * Return 1 if the task is a registered audit daemon, 0 otherwise.
217  */
218 int auditd_test_task(struct task_struct *task)
219 {
220 	int rc;
221 	struct auditd_connection *ac;
222 
223 	rcu_read_lock();
224 	ac = rcu_dereference(auditd_conn);
225 	rc = (ac && ac->pid == task_tgid(task) ? 1 : 0);
226 	rcu_read_unlock();
227 
228 	return rc;
229 }
230 
231 /**
232  * audit_ctl_lock - Take the audit control lock
233  */
234 void audit_ctl_lock(void)
235 {
236 	mutex_lock(&audit_cmd_mutex.lock);
237 	audit_cmd_mutex.owner = current;
238 }
239 
240 /**
241  * audit_ctl_unlock - Drop the audit control lock
242  */
243 void audit_ctl_unlock(void)
244 {
245 	audit_cmd_mutex.owner = NULL;
246 	mutex_unlock(&audit_cmd_mutex.lock);
247 }
248 
249 /**
250  * audit_ctl_owner_current - Test to see if the current task owns the lock
251  *
252  * Description:
253  * Return true if the current task owns the audit control lock, false if it
254  * doesn't own the lock.
255  */
256 static bool audit_ctl_owner_current(void)
257 {
258 	return (current == audit_cmd_mutex.owner);
259 }
260 
261 /**
262  * auditd_pid_vnr - Return the auditd PID relative to the namespace
263  *
264  * Description:
265  * Returns the PID in relation to the namespace, 0 on failure.
266  */
267 static pid_t auditd_pid_vnr(void)
268 {
269 	pid_t pid;
270 	const struct auditd_connection *ac;
271 
272 	rcu_read_lock();
273 	ac = rcu_dereference(auditd_conn);
274 	if (!ac || !ac->pid)
275 		pid = 0;
276 	else
277 		pid = pid_vnr(ac->pid);
278 	rcu_read_unlock();
279 
280 	return pid;
281 }
282 
283 /**
284  * audit_get_sk - Return the audit socket for the given network namespace
285  * @net: the destination network namespace
286  *
287  * Description:
288  * Returns the sock pointer if valid, NULL otherwise.  The caller must ensure
289  * that a reference is held for the network namespace while the sock is in use.
290  */
291 static struct sock *audit_get_sk(const struct net *net)
292 {
293 	struct audit_net *aunet;
294 
295 	if (!net)
296 		return NULL;
297 
298 	aunet = net_generic(net, audit_net_id);
299 	return aunet->sk;
300 }
301 
302 void audit_panic(const char *message)
303 {
304 	switch (audit_failure) {
305 	case AUDIT_FAIL_SILENT:
306 		break;
307 	case AUDIT_FAIL_PRINTK:
308 		if (printk_ratelimit())
309 			pr_err("%s\n", message);
310 		break;
311 	case AUDIT_FAIL_PANIC:
312 		panic("audit: %s\n", message);
313 		break;
314 	}
315 }
316 
317 static inline int audit_rate_check(void)
318 {
319 	static unsigned long	last_check = 0;
320 	static int		messages   = 0;
321 	static DEFINE_SPINLOCK(lock);
322 	unsigned long		flags;
323 	unsigned long		now;
324 	unsigned long		elapsed;
325 	int			retval	   = 0;
326 
327 	if (!audit_rate_limit) return 1;
328 
329 	spin_lock_irqsave(&lock, flags);
330 	if (++messages < audit_rate_limit) {
331 		retval = 1;
332 	} else {
333 		now     = jiffies;
334 		elapsed = now - last_check;
335 		if (elapsed > HZ) {
336 			last_check = now;
337 			messages   = 0;
338 			retval     = 1;
339 		}
340 	}
341 	spin_unlock_irqrestore(&lock, flags);
342 
343 	return retval;
344 }
345 
346 /**
347  * audit_log_lost - conditionally log lost audit message event
348  * @message: the message stating reason for lost audit message
349  *
350  * Emit at least 1 message per second, even if audit_rate_check is
351  * throttling.
352  * Always increment the lost messages counter.
353 */
354 void audit_log_lost(const char *message)
355 {
356 	static unsigned long	last_msg = 0;
357 	static DEFINE_SPINLOCK(lock);
358 	unsigned long		flags;
359 	unsigned long		now;
360 	int			print;
361 
362 	atomic_inc(&audit_lost);
363 
364 	print = (audit_failure == AUDIT_FAIL_PANIC || !audit_rate_limit);
365 
366 	if (!print) {
367 		spin_lock_irqsave(&lock, flags);
368 		now = jiffies;
369 		if (now - last_msg > HZ) {
370 			print = 1;
371 			last_msg = now;
372 		}
373 		spin_unlock_irqrestore(&lock, flags);
374 	}
375 
376 	if (print) {
377 		if (printk_ratelimit())
378 			pr_warn("audit_lost=%u audit_rate_limit=%u audit_backlog_limit=%u\n",
379 				atomic_read(&audit_lost),
380 				audit_rate_limit,
381 				audit_backlog_limit);
382 		audit_panic(message);
383 	}
384 }
385 
386 static int audit_log_config_change(char *function_name, u32 new, u32 old,
387 				   int allow_changes)
388 {
389 	struct audit_buffer *ab;
390 	int rc = 0;
391 
392 	ab = audit_log_start(audit_context(), GFP_KERNEL, AUDIT_CONFIG_CHANGE);
393 	if (unlikely(!ab))
394 		return rc;
395 	audit_log_format(ab, "op=set %s=%u old=%u ", function_name, new, old);
396 	audit_log_session_info(ab);
397 	rc = audit_log_task_context(ab);
398 	if (rc)
399 		allow_changes = 0; /* Something weird, deny request */
400 	audit_log_format(ab, " res=%d", allow_changes);
401 	audit_log_end(ab);
402 	return rc;
403 }
404 
405 static int audit_do_config_change(char *function_name, u32 *to_change, u32 new)
406 {
407 	int allow_changes, rc = 0;
408 	u32 old = *to_change;
409 
410 	/* check if we are locked */
411 	if (audit_enabled == AUDIT_LOCKED)
412 		allow_changes = 0;
413 	else
414 		allow_changes = 1;
415 
416 	if (audit_enabled != AUDIT_OFF) {
417 		rc = audit_log_config_change(function_name, new, old, allow_changes);
418 		if (rc)
419 			allow_changes = 0;
420 	}
421 
422 	/* If we are allowed, make the change */
423 	if (allow_changes == 1)
424 		*to_change = new;
425 	/* Not allowed, update reason */
426 	else if (rc == 0)
427 		rc = -EPERM;
428 	return rc;
429 }
430 
431 static int audit_set_rate_limit(u32 limit)
432 {
433 	return audit_do_config_change("audit_rate_limit", &audit_rate_limit, limit);
434 }
435 
436 static int audit_set_backlog_limit(u32 limit)
437 {
438 	return audit_do_config_change("audit_backlog_limit", &audit_backlog_limit, limit);
439 }
440 
441 static int audit_set_backlog_wait_time(u32 timeout)
442 {
443 	return audit_do_config_change("audit_backlog_wait_time",
444 				      &audit_backlog_wait_time, timeout);
445 }
446 
447 static int audit_set_enabled(u32 state)
448 {
449 	int rc;
450 	if (state > AUDIT_LOCKED)
451 		return -EINVAL;
452 
453 	rc =  audit_do_config_change("audit_enabled", &audit_enabled, state);
454 	if (!rc)
455 		audit_ever_enabled |= !!state;
456 
457 	return rc;
458 }
459 
460 static int audit_set_failure(u32 state)
461 {
462 	if (state != AUDIT_FAIL_SILENT
463 	    && state != AUDIT_FAIL_PRINTK
464 	    && state != AUDIT_FAIL_PANIC)
465 		return -EINVAL;
466 
467 	return audit_do_config_change("audit_failure", &audit_failure, state);
468 }
469 
470 /**
471  * auditd_conn_free - RCU helper to release an auditd connection struct
472  * @rcu: RCU head
473  *
474  * Description:
475  * Drop any references inside the auditd connection tracking struct and free
476  * the memory.
477  */
478 static void auditd_conn_free(struct rcu_head *rcu)
479 {
480 	struct auditd_connection *ac;
481 
482 	ac = container_of(rcu, struct auditd_connection, rcu);
483 	put_pid(ac->pid);
484 	put_net(ac->net);
485 	kfree(ac);
486 }
487 
488 /**
489  * auditd_set - Set/Reset the auditd connection state
490  * @pid: auditd PID
491  * @portid: auditd netlink portid
492  * @net: auditd network namespace pointer
493  *
494  * Description:
495  * This function will obtain and drop network namespace references as
496  * necessary.  Returns zero on success, negative values on failure.
497  */
498 static int auditd_set(struct pid *pid, u32 portid, struct net *net)
499 {
500 	unsigned long flags;
501 	struct auditd_connection *ac_old, *ac_new;
502 
503 	if (!pid || !net)
504 		return -EINVAL;
505 
506 	ac_new = kzalloc(sizeof(*ac_new), GFP_KERNEL);
507 	if (!ac_new)
508 		return -ENOMEM;
509 	ac_new->pid = get_pid(pid);
510 	ac_new->portid = portid;
511 	ac_new->net = get_net(net);
512 
513 	spin_lock_irqsave(&auditd_conn_lock, flags);
514 	ac_old = rcu_dereference_protected(auditd_conn,
515 					   lockdep_is_held(&auditd_conn_lock));
516 	rcu_assign_pointer(auditd_conn, ac_new);
517 	spin_unlock_irqrestore(&auditd_conn_lock, flags);
518 
519 	if (ac_old)
520 		call_rcu(&ac_old->rcu, auditd_conn_free);
521 
522 	return 0;
523 }
524 
525 /**
526  * kauditd_printk_skb - Print the audit record to the ring buffer
527  * @skb: audit record
528  *
529  * Whatever the reason, this packet may not make it to the auditd connection
530  * so write it via printk so the information isn't completely lost.
531  */
532 static void kauditd_printk_skb(struct sk_buff *skb)
533 {
534 	struct nlmsghdr *nlh = nlmsg_hdr(skb);
535 	char *data = nlmsg_data(nlh);
536 
537 	if (nlh->nlmsg_type != AUDIT_EOE && printk_ratelimit())
538 		pr_notice("type=%d %s\n", nlh->nlmsg_type, data);
539 }
540 
541 /**
542  * kauditd_rehold_skb - Handle a audit record send failure in the hold queue
543  * @skb: audit record
544  *
545  * Description:
546  * This should only be used by the kauditd_thread when it fails to flush the
547  * hold queue.
548  */
549 static void kauditd_rehold_skb(struct sk_buff *skb)
550 {
551 	/* put the record back in the queue at the same place */
552 	skb_queue_head(&audit_hold_queue, skb);
553 }
554 
555 /**
556  * kauditd_hold_skb - Queue an audit record, waiting for auditd
557  * @skb: audit record
558  *
559  * Description:
560  * Queue the audit record, waiting for an instance of auditd.  When this
561  * function is called we haven't given up yet on sending the record, but things
562  * are not looking good.  The first thing we want to do is try to write the
563  * record via printk and then see if we want to try and hold on to the record
564  * and queue it, if we have room.  If we want to hold on to the record, but we
565  * don't have room, record a record lost message.
566  */
567 static void kauditd_hold_skb(struct sk_buff *skb)
568 {
569 	/* at this point it is uncertain if we will ever send this to auditd so
570 	 * try to send the message via printk before we go any further */
571 	kauditd_printk_skb(skb);
572 
573 	/* can we just silently drop the message? */
574 	if (!audit_default) {
575 		kfree_skb(skb);
576 		return;
577 	}
578 
579 	/* if we have room, queue the message */
580 	if (!audit_backlog_limit ||
581 	    skb_queue_len(&audit_hold_queue) < audit_backlog_limit) {
582 		skb_queue_tail(&audit_hold_queue, skb);
583 		return;
584 	}
585 
586 	/* we have no other options - drop the message */
587 	audit_log_lost("kauditd hold queue overflow");
588 	kfree_skb(skb);
589 }
590 
591 /**
592  * kauditd_retry_skb - Queue an audit record, attempt to send again to auditd
593  * @skb: audit record
594  *
595  * Description:
596  * Not as serious as kauditd_hold_skb() as we still have a connected auditd,
597  * but for some reason we are having problems sending it audit records so
598  * queue the given record and attempt to resend.
599  */
600 static void kauditd_retry_skb(struct sk_buff *skb)
601 {
602 	/* NOTE: because records should only live in the retry queue for a
603 	 * short period of time, before either being sent or moved to the hold
604 	 * queue, we don't currently enforce a limit on this queue */
605 	skb_queue_tail(&audit_retry_queue, skb);
606 }
607 
608 /**
609  * auditd_reset - Disconnect the auditd connection
610  * @ac: auditd connection state
611  *
612  * Description:
613  * Break the auditd/kauditd connection and move all the queued records into the
614  * hold queue in case auditd reconnects.  It is important to note that the @ac
615  * pointer should never be dereferenced inside this function as it may be NULL
616  * or invalid, you can only compare the memory address!  If @ac is NULL then
617  * the connection will always be reset.
618  */
619 static void auditd_reset(const struct auditd_connection *ac)
620 {
621 	unsigned long flags;
622 	struct sk_buff *skb;
623 	struct auditd_connection *ac_old;
624 
625 	/* if it isn't already broken, break the connection */
626 	spin_lock_irqsave(&auditd_conn_lock, flags);
627 	ac_old = rcu_dereference_protected(auditd_conn,
628 					   lockdep_is_held(&auditd_conn_lock));
629 	if (ac && ac != ac_old) {
630 		/* someone already registered a new auditd connection */
631 		spin_unlock_irqrestore(&auditd_conn_lock, flags);
632 		return;
633 	}
634 	rcu_assign_pointer(auditd_conn, NULL);
635 	spin_unlock_irqrestore(&auditd_conn_lock, flags);
636 
637 	if (ac_old)
638 		call_rcu(&ac_old->rcu, auditd_conn_free);
639 
640 	/* flush the retry queue to the hold queue, but don't touch the main
641 	 * queue since we need to process that normally for multicast */
642 	while ((skb = skb_dequeue(&audit_retry_queue)))
643 		kauditd_hold_skb(skb);
644 }
645 
646 /**
647  * auditd_send_unicast_skb - Send a record via unicast to auditd
648  * @skb: audit record
649  *
650  * Description:
651  * Send a skb to the audit daemon, returns positive/zero values on success and
652  * negative values on failure; in all cases the skb will be consumed by this
653  * function.  If the send results in -ECONNREFUSED the connection with auditd
654  * will be reset.  This function may sleep so callers should not hold any locks
655  * where this would cause a problem.
656  */
657 static int auditd_send_unicast_skb(struct sk_buff *skb)
658 {
659 	int rc;
660 	u32 portid;
661 	struct net *net;
662 	struct sock *sk;
663 	struct auditd_connection *ac;
664 
665 	/* NOTE: we can't call netlink_unicast while in the RCU section so
666 	 *       take a reference to the network namespace and grab local
667 	 *       copies of the namespace, the sock, and the portid; the
668 	 *       namespace and sock aren't going to go away while we hold a
669 	 *       reference and if the portid does become invalid after the RCU
670 	 *       section netlink_unicast() should safely return an error */
671 
672 	rcu_read_lock();
673 	ac = rcu_dereference(auditd_conn);
674 	if (!ac) {
675 		rcu_read_unlock();
676 		kfree_skb(skb);
677 		rc = -ECONNREFUSED;
678 		goto err;
679 	}
680 	net = get_net(ac->net);
681 	sk = audit_get_sk(net);
682 	portid = ac->portid;
683 	rcu_read_unlock();
684 
685 	rc = netlink_unicast(sk, skb, portid, 0);
686 	put_net(net);
687 	if (rc < 0)
688 		goto err;
689 
690 	return rc;
691 
692 err:
693 	if (ac && rc == -ECONNREFUSED)
694 		auditd_reset(ac);
695 	return rc;
696 }
697 
698 /**
699  * kauditd_send_queue - Helper for kauditd_thread to flush skb queues
700  * @sk: the sending sock
701  * @portid: the netlink destination
702  * @queue: the skb queue to process
703  * @retry_limit: limit on number of netlink unicast failures
704  * @skb_hook: per-skb hook for additional processing
705  * @err_hook: hook called if the skb fails the netlink unicast send
706  *
707  * Description:
708  * Run through the given queue and attempt to send the audit records to auditd,
709  * returns zero on success, negative values on failure.  It is up to the caller
710  * to ensure that the @sk is valid for the duration of this function.
711  *
712  */
713 static int kauditd_send_queue(struct sock *sk, u32 portid,
714 			      struct sk_buff_head *queue,
715 			      unsigned int retry_limit,
716 			      void (*skb_hook)(struct sk_buff *skb),
717 			      void (*err_hook)(struct sk_buff *skb))
718 {
719 	int rc = 0;
720 	struct sk_buff *skb;
721 	static unsigned int failed = 0;
722 
723 	/* NOTE: kauditd_thread takes care of all our locking, we just use
724 	 *       the netlink info passed to us (e.g. sk and portid) */
725 
726 	while ((skb = skb_dequeue(queue))) {
727 		/* call the skb_hook for each skb we touch */
728 		if (skb_hook)
729 			(*skb_hook)(skb);
730 
731 		/* can we send to anyone via unicast? */
732 		if (!sk) {
733 			if (err_hook)
734 				(*err_hook)(skb);
735 			continue;
736 		}
737 
738 		/* grab an extra skb reference in case of error */
739 		skb_get(skb);
740 		rc = netlink_unicast(sk, skb, portid, 0);
741 		if (rc < 0) {
742 			/* fatal failure for our queue flush attempt? */
743 			if (++failed >= retry_limit ||
744 			    rc == -ECONNREFUSED || rc == -EPERM) {
745 				/* yes - error processing for the queue */
746 				sk = NULL;
747 				if (err_hook)
748 					(*err_hook)(skb);
749 				if (!skb_hook)
750 					goto out;
751 				/* keep processing with the skb_hook */
752 				continue;
753 			} else
754 				/* no - requeue to preserve ordering */
755 				skb_queue_head(queue, skb);
756 		} else {
757 			/* it worked - drop the extra reference and continue */
758 			consume_skb(skb);
759 			failed = 0;
760 		}
761 	}
762 
763 out:
764 	return (rc >= 0 ? 0 : rc);
765 }
766 
767 /*
768  * kauditd_send_multicast_skb - Send a record to any multicast listeners
769  * @skb: audit record
770  *
771  * Description:
772  * Write a multicast message to anyone listening in the initial network
773  * namespace.  This function doesn't consume an skb as might be expected since
774  * it has to copy it anyways.
775  */
776 static void kauditd_send_multicast_skb(struct sk_buff *skb)
777 {
778 	struct sk_buff *copy;
779 	struct sock *sock = audit_get_sk(&init_net);
780 	struct nlmsghdr *nlh;
781 
782 	/* NOTE: we are not taking an additional reference for init_net since
783 	 *       we don't have to worry about it going away */
784 
785 	if (!netlink_has_listeners(sock, AUDIT_NLGRP_READLOG))
786 		return;
787 
788 	/*
789 	 * The seemingly wasteful skb_copy() rather than bumping the refcount
790 	 * using skb_get() is necessary because non-standard mods are made to
791 	 * the skb by the original kaudit unicast socket send routine.  The
792 	 * existing auditd daemon assumes this breakage.  Fixing this would
793 	 * require co-ordinating a change in the established protocol between
794 	 * the kaudit kernel subsystem and the auditd userspace code.  There is
795 	 * no reason for new multicast clients to continue with this
796 	 * non-compliance.
797 	 */
798 	copy = skb_copy(skb, GFP_KERNEL);
799 	if (!copy)
800 		return;
801 	nlh = nlmsg_hdr(copy);
802 	nlh->nlmsg_len = skb->len;
803 
804 	nlmsg_multicast(sock, copy, 0, AUDIT_NLGRP_READLOG, GFP_KERNEL);
805 }
806 
807 /**
808  * kauditd_thread - Worker thread to send audit records to userspace
809  * @dummy: unused
810  */
811 static int kauditd_thread(void *dummy)
812 {
813 	int rc;
814 	u32 portid = 0;
815 	struct net *net = NULL;
816 	struct sock *sk = NULL;
817 	struct auditd_connection *ac;
818 
819 #define UNICAST_RETRIES 5
820 
821 	set_freezable();
822 	while (!kthread_should_stop()) {
823 		/* NOTE: see the lock comments in auditd_send_unicast_skb() */
824 		rcu_read_lock();
825 		ac = rcu_dereference(auditd_conn);
826 		if (!ac) {
827 			rcu_read_unlock();
828 			goto main_queue;
829 		}
830 		net = get_net(ac->net);
831 		sk = audit_get_sk(net);
832 		portid = ac->portid;
833 		rcu_read_unlock();
834 
835 		/* attempt to flush the hold queue */
836 		rc = kauditd_send_queue(sk, portid,
837 					&audit_hold_queue, UNICAST_RETRIES,
838 					NULL, kauditd_rehold_skb);
839 		if (rc < 0) {
840 			sk = NULL;
841 			auditd_reset(ac);
842 			goto main_queue;
843 		}
844 
845 		/* attempt to flush the retry queue */
846 		rc = kauditd_send_queue(sk, portid,
847 					&audit_retry_queue, UNICAST_RETRIES,
848 					NULL, kauditd_hold_skb);
849 		if (rc < 0) {
850 			sk = NULL;
851 			auditd_reset(ac);
852 			goto main_queue;
853 		}
854 
855 main_queue:
856 		/* process the main queue - do the multicast send and attempt
857 		 * unicast, dump failed record sends to the retry queue; if
858 		 * sk == NULL due to previous failures we will just do the
859 		 * multicast send and move the record to the hold queue */
860 		rc = kauditd_send_queue(sk, portid, &audit_queue, 1,
861 					kauditd_send_multicast_skb,
862 					(sk ?
863 					 kauditd_retry_skb : kauditd_hold_skb));
864 		if (ac && rc < 0)
865 			auditd_reset(ac);
866 		sk = NULL;
867 
868 		/* drop our netns reference, no auditd sends past this line */
869 		if (net) {
870 			put_net(net);
871 			net = NULL;
872 		}
873 
874 		/* we have processed all the queues so wake everyone */
875 		wake_up(&audit_backlog_wait);
876 
877 		/* NOTE: we want to wake up if there is anything on the queue,
878 		 *       regardless of if an auditd is connected, as we need to
879 		 *       do the multicast send and rotate records from the
880 		 *       main queue to the retry/hold queues */
881 		wait_event_freezable(kauditd_wait,
882 				     (skb_queue_len(&audit_queue) ? 1 : 0));
883 	}
884 
885 	return 0;
886 }
887 
888 int audit_send_list_thread(void *_dest)
889 {
890 	struct audit_netlink_list *dest = _dest;
891 	struct sk_buff *skb;
892 	struct sock *sk = audit_get_sk(dest->net);
893 
894 	/* wait for parent to finish and send an ACK */
895 	audit_ctl_lock();
896 	audit_ctl_unlock();
897 
898 	while ((skb = __skb_dequeue(&dest->q)) != NULL)
899 		netlink_unicast(sk, skb, dest->portid, 0);
900 
901 	put_net(dest->net);
902 	kfree(dest);
903 
904 	return 0;
905 }
906 
907 struct sk_buff *audit_make_reply(int seq, int type, int done,
908 				 int multi, const void *payload, int size)
909 {
910 	struct sk_buff	*skb;
911 	struct nlmsghdr	*nlh;
912 	void		*data;
913 	int		flags = multi ? NLM_F_MULTI : 0;
914 	int		t     = done  ? NLMSG_DONE  : type;
915 
916 	skb = nlmsg_new(size, GFP_KERNEL);
917 	if (!skb)
918 		return NULL;
919 
920 	nlh	= nlmsg_put(skb, 0, seq, t, size, flags);
921 	if (!nlh)
922 		goto out_kfree_skb;
923 	data = nlmsg_data(nlh);
924 	memcpy(data, payload, size);
925 	return skb;
926 
927 out_kfree_skb:
928 	kfree_skb(skb);
929 	return NULL;
930 }
931 
932 static void audit_free_reply(struct audit_reply *reply)
933 {
934 	if (!reply)
935 		return;
936 
937 	kfree_skb(reply->skb);
938 	if (reply->net)
939 		put_net(reply->net);
940 	kfree(reply);
941 }
942 
943 static int audit_send_reply_thread(void *arg)
944 {
945 	struct audit_reply *reply = (struct audit_reply *)arg;
946 
947 	audit_ctl_lock();
948 	audit_ctl_unlock();
949 
950 	/* Ignore failure. It'll only happen if the sender goes away,
951 	   because our timeout is set to infinite. */
952 	netlink_unicast(audit_get_sk(reply->net), reply->skb, reply->portid, 0);
953 	reply->skb = NULL;
954 	audit_free_reply(reply);
955 	return 0;
956 }
957 
958 /**
959  * audit_send_reply - send an audit reply message via netlink
960  * @request_skb: skb of request we are replying to (used to target the reply)
961  * @seq: sequence number
962  * @type: audit message type
963  * @done: done (last) flag
964  * @multi: multi-part message flag
965  * @payload: payload data
966  * @size: payload size
967  *
968  * Allocates a skb, builds the netlink message, and sends it to the port id.
969  */
970 static void audit_send_reply(struct sk_buff *request_skb, int seq, int type, int done,
971 			     int multi, const void *payload, int size)
972 {
973 	struct task_struct *tsk;
974 	struct audit_reply *reply;
975 
976 	reply = kzalloc(sizeof(*reply), GFP_KERNEL);
977 	if (!reply)
978 		return;
979 
980 	reply->skb = audit_make_reply(seq, type, done, multi, payload, size);
981 	if (!reply->skb)
982 		goto err;
983 	reply->net = get_net(sock_net(NETLINK_CB(request_skb).sk));
984 	reply->portid = NETLINK_CB(request_skb).portid;
985 
986 	tsk = kthread_run(audit_send_reply_thread, reply, "audit_send_reply");
987 	if (IS_ERR(tsk))
988 		goto err;
989 
990 	return;
991 
992 err:
993 	audit_free_reply(reply);
994 }
995 
996 /*
997  * Check for appropriate CAP_AUDIT_ capabilities on incoming audit
998  * control messages.
999  */
1000 static int audit_netlink_ok(struct sk_buff *skb, u16 msg_type)
1001 {
1002 	int err = 0;
1003 
1004 	/* Only support initial user namespace for now. */
1005 	/*
1006 	 * We return ECONNREFUSED because it tricks userspace into thinking
1007 	 * that audit was not configured into the kernel.  Lots of users
1008 	 * configure their PAM stack (because that's what the distro does)
1009 	 * to reject login if unable to send messages to audit.  If we return
1010 	 * ECONNREFUSED the PAM stack thinks the kernel does not have audit
1011 	 * configured in and will let login proceed.  If we return EPERM
1012 	 * userspace will reject all logins.  This should be removed when we
1013 	 * support non init namespaces!!
1014 	 */
1015 	if (current_user_ns() != &init_user_ns)
1016 		return -ECONNREFUSED;
1017 
1018 	switch (msg_type) {
1019 	case AUDIT_LIST:
1020 	case AUDIT_ADD:
1021 	case AUDIT_DEL:
1022 		return -EOPNOTSUPP;
1023 	case AUDIT_GET:
1024 	case AUDIT_SET:
1025 	case AUDIT_GET_FEATURE:
1026 	case AUDIT_SET_FEATURE:
1027 	case AUDIT_LIST_RULES:
1028 	case AUDIT_ADD_RULE:
1029 	case AUDIT_DEL_RULE:
1030 	case AUDIT_SIGNAL_INFO:
1031 	case AUDIT_TTY_GET:
1032 	case AUDIT_TTY_SET:
1033 	case AUDIT_TRIM:
1034 	case AUDIT_MAKE_EQUIV:
1035 		/* Only support auditd and auditctl in initial pid namespace
1036 		 * for now. */
1037 		if (task_active_pid_ns(current) != &init_pid_ns)
1038 			return -EPERM;
1039 
1040 		if (!netlink_capable(skb, CAP_AUDIT_CONTROL))
1041 			err = -EPERM;
1042 		break;
1043 	case AUDIT_USER:
1044 	case AUDIT_FIRST_USER_MSG ... AUDIT_LAST_USER_MSG:
1045 	case AUDIT_FIRST_USER_MSG2 ... AUDIT_LAST_USER_MSG2:
1046 		if (!netlink_capable(skb, CAP_AUDIT_WRITE))
1047 			err = -EPERM;
1048 		break;
1049 	default:  /* bad msg */
1050 		err = -EINVAL;
1051 	}
1052 
1053 	return err;
1054 }
1055 
1056 static void audit_log_common_recv_msg(struct audit_context *context,
1057 					struct audit_buffer **ab, u16 msg_type)
1058 {
1059 	uid_t uid = from_kuid(&init_user_ns, current_uid());
1060 	pid_t pid = task_tgid_nr(current);
1061 
1062 	if (!audit_enabled && msg_type != AUDIT_USER_AVC) {
1063 		*ab = NULL;
1064 		return;
1065 	}
1066 
1067 	*ab = audit_log_start(context, GFP_KERNEL, msg_type);
1068 	if (unlikely(!*ab))
1069 		return;
1070 	audit_log_format(*ab, "pid=%d uid=%u ", pid, uid);
1071 	audit_log_session_info(*ab);
1072 	audit_log_task_context(*ab);
1073 }
1074 
1075 static inline void audit_log_user_recv_msg(struct audit_buffer **ab,
1076 					   u16 msg_type)
1077 {
1078 	audit_log_common_recv_msg(NULL, ab, msg_type);
1079 }
1080 
1081 int is_audit_feature_set(int i)
1082 {
1083 	return af.features & AUDIT_FEATURE_TO_MASK(i);
1084 }
1085 
1086 
1087 static int audit_get_feature(struct sk_buff *skb)
1088 {
1089 	u32 seq;
1090 
1091 	seq = nlmsg_hdr(skb)->nlmsg_seq;
1092 
1093 	audit_send_reply(skb, seq, AUDIT_GET_FEATURE, 0, 0, &af, sizeof(af));
1094 
1095 	return 0;
1096 }
1097 
1098 static void audit_log_feature_change(int which, u32 old_feature, u32 new_feature,
1099 				     u32 old_lock, u32 new_lock, int res)
1100 {
1101 	struct audit_buffer *ab;
1102 
1103 	if (audit_enabled == AUDIT_OFF)
1104 		return;
1105 
1106 	ab = audit_log_start(audit_context(), GFP_KERNEL, AUDIT_FEATURE_CHANGE);
1107 	if (!ab)
1108 		return;
1109 	audit_log_task_info(ab);
1110 	audit_log_format(ab, " feature=%s old=%u new=%u old_lock=%u new_lock=%u res=%d",
1111 			 audit_feature_names[which], !!old_feature, !!new_feature,
1112 			 !!old_lock, !!new_lock, res);
1113 	audit_log_end(ab);
1114 }
1115 
1116 static int audit_set_feature(struct audit_features *uaf)
1117 {
1118 	int i;
1119 
1120 	BUILD_BUG_ON(AUDIT_LAST_FEATURE + 1 > ARRAY_SIZE(audit_feature_names));
1121 
1122 	/* if there is ever a version 2 we should handle that here */
1123 
1124 	for (i = 0; i <= AUDIT_LAST_FEATURE; i++) {
1125 		u32 feature = AUDIT_FEATURE_TO_MASK(i);
1126 		u32 old_feature, new_feature, old_lock, new_lock;
1127 
1128 		/* if we are not changing this feature, move along */
1129 		if (!(feature & uaf->mask))
1130 			continue;
1131 
1132 		old_feature = af.features & feature;
1133 		new_feature = uaf->features & feature;
1134 		new_lock = (uaf->lock | af.lock) & feature;
1135 		old_lock = af.lock & feature;
1136 
1137 		/* are we changing a locked feature? */
1138 		if (old_lock && (new_feature != old_feature)) {
1139 			audit_log_feature_change(i, old_feature, new_feature,
1140 						 old_lock, new_lock, 0);
1141 			return -EPERM;
1142 		}
1143 	}
1144 	/* nothing invalid, do the changes */
1145 	for (i = 0; i <= AUDIT_LAST_FEATURE; i++) {
1146 		u32 feature = AUDIT_FEATURE_TO_MASK(i);
1147 		u32 old_feature, new_feature, old_lock, new_lock;
1148 
1149 		/* if we are not changing this feature, move along */
1150 		if (!(feature & uaf->mask))
1151 			continue;
1152 
1153 		old_feature = af.features & feature;
1154 		new_feature = uaf->features & feature;
1155 		old_lock = af.lock & feature;
1156 		new_lock = (uaf->lock | af.lock) & feature;
1157 
1158 		if (new_feature != old_feature)
1159 			audit_log_feature_change(i, old_feature, new_feature,
1160 						 old_lock, new_lock, 1);
1161 
1162 		if (new_feature)
1163 			af.features |= feature;
1164 		else
1165 			af.features &= ~feature;
1166 		af.lock |= new_lock;
1167 	}
1168 
1169 	return 0;
1170 }
1171 
1172 static int audit_replace(struct pid *pid)
1173 {
1174 	pid_t pvnr;
1175 	struct sk_buff *skb;
1176 
1177 	pvnr = pid_vnr(pid);
1178 	skb = audit_make_reply(0, AUDIT_REPLACE, 0, 0, &pvnr, sizeof(pvnr));
1179 	if (!skb)
1180 		return -ENOMEM;
1181 	return auditd_send_unicast_skb(skb);
1182 }
1183 
1184 static int audit_receive_msg(struct sk_buff *skb, struct nlmsghdr *nlh)
1185 {
1186 	u32			seq;
1187 	void			*data;
1188 	int			data_len;
1189 	int			err;
1190 	struct audit_buffer	*ab;
1191 	u16			msg_type = nlh->nlmsg_type;
1192 	struct audit_sig_info   *sig_data;
1193 	char			*ctx = NULL;
1194 	u32			len;
1195 
1196 	err = audit_netlink_ok(skb, msg_type);
1197 	if (err)
1198 		return err;
1199 
1200 	seq  = nlh->nlmsg_seq;
1201 	data = nlmsg_data(nlh);
1202 	data_len = nlmsg_len(nlh);
1203 
1204 	switch (msg_type) {
1205 	case AUDIT_GET: {
1206 		struct audit_status	s;
1207 		memset(&s, 0, sizeof(s));
1208 		s.enabled		   = audit_enabled;
1209 		s.failure		   = audit_failure;
1210 		/* NOTE: use pid_vnr() so the PID is relative to the current
1211 		 *       namespace */
1212 		s.pid			   = auditd_pid_vnr();
1213 		s.rate_limit		   = audit_rate_limit;
1214 		s.backlog_limit		   = audit_backlog_limit;
1215 		s.lost			   = atomic_read(&audit_lost);
1216 		s.backlog		   = skb_queue_len(&audit_queue);
1217 		s.feature_bitmap	   = AUDIT_FEATURE_BITMAP_ALL;
1218 		s.backlog_wait_time	   = audit_backlog_wait_time;
1219 		s.backlog_wait_time_actual = atomic_read(&audit_backlog_wait_time_actual);
1220 		audit_send_reply(skb, seq, AUDIT_GET, 0, 0, &s, sizeof(s));
1221 		break;
1222 	}
1223 	case AUDIT_SET: {
1224 		struct audit_status	s;
1225 		memset(&s, 0, sizeof(s));
1226 		/* guard against past and future API changes */
1227 		memcpy(&s, data, min_t(size_t, sizeof(s), data_len));
1228 		if (s.mask & AUDIT_STATUS_ENABLED) {
1229 			err = audit_set_enabled(s.enabled);
1230 			if (err < 0)
1231 				return err;
1232 		}
1233 		if (s.mask & AUDIT_STATUS_FAILURE) {
1234 			err = audit_set_failure(s.failure);
1235 			if (err < 0)
1236 				return err;
1237 		}
1238 		if (s.mask & AUDIT_STATUS_PID) {
1239 			/* NOTE: we are using the vnr PID functions below
1240 			 *       because the s.pid value is relative to the
1241 			 *       namespace of the caller; at present this
1242 			 *       doesn't matter much since you can really only
1243 			 *       run auditd from the initial pid namespace, but
1244 			 *       something to keep in mind if this changes */
1245 			pid_t new_pid = s.pid;
1246 			pid_t auditd_pid;
1247 			struct pid *req_pid = task_tgid(current);
1248 
1249 			/* Sanity check - PID values must match. Setting
1250 			 * pid to 0 is how auditd ends auditing. */
1251 			if (new_pid && (new_pid != pid_vnr(req_pid)))
1252 				return -EINVAL;
1253 
1254 			/* test the auditd connection */
1255 			audit_replace(req_pid);
1256 
1257 			auditd_pid = auditd_pid_vnr();
1258 			if (auditd_pid) {
1259 				/* replacing a healthy auditd is not allowed */
1260 				if (new_pid) {
1261 					audit_log_config_change("audit_pid",
1262 							new_pid, auditd_pid, 0);
1263 					return -EEXIST;
1264 				}
1265 				/* only current auditd can unregister itself */
1266 				if (pid_vnr(req_pid) != auditd_pid) {
1267 					audit_log_config_change("audit_pid",
1268 							new_pid, auditd_pid, 0);
1269 					return -EACCES;
1270 				}
1271 			}
1272 
1273 			if (new_pid) {
1274 				/* register a new auditd connection */
1275 				err = auditd_set(req_pid,
1276 						 NETLINK_CB(skb).portid,
1277 						 sock_net(NETLINK_CB(skb).sk));
1278 				if (audit_enabled != AUDIT_OFF)
1279 					audit_log_config_change("audit_pid",
1280 								new_pid,
1281 								auditd_pid,
1282 								err ? 0 : 1);
1283 				if (err)
1284 					return err;
1285 
1286 				/* try to process any backlog */
1287 				wake_up_interruptible(&kauditd_wait);
1288 			} else {
1289 				if (audit_enabled != AUDIT_OFF)
1290 					audit_log_config_change("audit_pid",
1291 								new_pid,
1292 								auditd_pid, 1);
1293 
1294 				/* unregister the auditd connection */
1295 				auditd_reset(NULL);
1296 			}
1297 		}
1298 		if (s.mask & AUDIT_STATUS_RATE_LIMIT) {
1299 			err = audit_set_rate_limit(s.rate_limit);
1300 			if (err < 0)
1301 				return err;
1302 		}
1303 		if (s.mask & AUDIT_STATUS_BACKLOG_LIMIT) {
1304 			err = audit_set_backlog_limit(s.backlog_limit);
1305 			if (err < 0)
1306 				return err;
1307 		}
1308 		if (s.mask & AUDIT_STATUS_BACKLOG_WAIT_TIME) {
1309 			if (sizeof(s) > (size_t)nlh->nlmsg_len)
1310 				return -EINVAL;
1311 			if (s.backlog_wait_time > 10*AUDIT_BACKLOG_WAIT_TIME)
1312 				return -EINVAL;
1313 			err = audit_set_backlog_wait_time(s.backlog_wait_time);
1314 			if (err < 0)
1315 				return err;
1316 		}
1317 		if (s.mask == AUDIT_STATUS_LOST) {
1318 			u32 lost = atomic_xchg(&audit_lost, 0);
1319 
1320 			audit_log_config_change("lost", 0, lost, 1);
1321 			return lost;
1322 		}
1323 		if (s.mask == AUDIT_STATUS_BACKLOG_WAIT_TIME_ACTUAL) {
1324 			u32 actual = atomic_xchg(&audit_backlog_wait_time_actual, 0);
1325 
1326 			audit_log_config_change("backlog_wait_time_actual", 0, actual, 1);
1327 			return actual;
1328 		}
1329 		break;
1330 	}
1331 	case AUDIT_GET_FEATURE:
1332 		err = audit_get_feature(skb);
1333 		if (err)
1334 			return err;
1335 		break;
1336 	case AUDIT_SET_FEATURE:
1337 		if (data_len < sizeof(struct audit_features))
1338 			return -EINVAL;
1339 		err = audit_set_feature(data);
1340 		if (err)
1341 			return err;
1342 		break;
1343 	case AUDIT_USER:
1344 	case AUDIT_FIRST_USER_MSG ... AUDIT_LAST_USER_MSG:
1345 	case AUDIT_FIRST_USER_MSG2 ... AUDIT_LAST_USER_MSG2:
1346 		if (!audit_enabled && msg_type != AUDIT_USER_AVC)
1347 			return 0;
1348 		/* exit early if there isn't at least one character to print */
1349 		if (data_len < 2)
1350 			return -EINVAL;
1351 
1352 		err = audit_filter(msg_type, AUDIT_FILTER_USER);
1353 		if (err == 1) { /* match or error */
1354 			char *str = data;
1355 
1356 			err = 0;
1357 			if (msg_type == AUDIT_USER_TTY) {
1358 				err = tty_audit_push();
1359 				if (err)
1360 					break;
1361 			}
1362 			audit_log_user_recv_msg(&ab, msg_type);
1363 			if (msg_type != AUDIT_USER_TTY) {
1364 				/* ensure NULL termination */
1365 				str[data_len - 1] = '\0';
1366 				audit_log_format(ab, " msg='%.*s'",
1367 						 AUDIT_MESSAGE_TEXT_MAX,
1368 						 str);
1369 			} else {
1370 				audit_log_format(ab, " data=");
1371 				if (data_len > 0 && str[data_len - 1] == '\0')
1372 					data_len--;
1373 				audit_log_n_untrustedstring(ab, str, data_len);
1374 			}
1375 			audit_log_end(ab);
1376 		}
1377 		break;
1378 	case AUDIT_ADD_RULE:
1379 	case AUDIT_DEL_RULE:
1380 		if (data_len < sizeof(struct audit_rule_data))
1381 			return -EINVAL;
1382 		if (audit_enabled == AUDIT_LOCKED) {
1383 			audit_log_common_recv_msg(audit_context(), &ab,
1384 						  AUDIT_CONFIG_CHANGE);
1385 			audit_log_format(ab, " op=%s audit_enabled=%d res=0",
1386 					 msg_type == AUDIT_ADD_RULE ?
1387 						"add_rule" : "remove_rule",
1388 					 audit_enabled);
1389 			audit_log_end(ab);
1390 			return -EPERM;
1391 		}
1392 		err = audit_rule_change(msg_type, seq, data, data_len);
1393 		break;
1394 	case AUDIT_LIST_RULES:
1395 		err = audit_list_rules_send(skb, seq);
1396 		break;
1397 	case AUDIT_TRIM:
1398 		audit_trim_trees();
1399 		audit_log_common_recv_msg(audit_context(), &ab,
1400 					  AUDIT_CONFIG_CHANGE);
1401 		audit_log_format(ab, " op=trim res=1");
1402 		audit_log_end(ab);
1403 		break;
1404 	case AUDIT_MAKE_EQUIV: {
1405 		void *bufp = data;
1406 		u32 sizes[2];
1407 		size_t msglen = data_len;
1408 		char *old, *new;
1409 
1410 		err = -EINVAL;
1411 		if (msglen < 2 * sizeof(u32))
1412 			break;
1413 		memcpy(sizes, bufp, 2 * sizeof(u32));
1414 		bufp += 2 * sizeof(u32);
1415 		msglen -= 2 * sizeof(u32);
1416 		old = audit_unpack_string(&bufp, &msglen, sizes[0]);
1417 		if (IS_ERR(old)) {
1418 			err = PTR_ERR(old);
1419 			break;
1420 		}
1421 		new = audit_unpack_string(&bufp, &msglen, sizes[1]);
1422 		if (IS_ERR(new)) {
1423 			err = PTR_ERR(new);
1424 			kfree(old);
1425 			break;
1426 		}
1427 		/* OK, here comes... */
1428 		err = audit_tag_tree(old, new);
1429 
1430 		audit_log_common_recv_msg(audit_context(), &ab,
1431 					  AUDIT_CONFIG_CHANGE);
1432 		audit_log_format(ab, " op=make_equiv old=");
1433 		audit_log_untrustedstring(ab, old);
1434 		audit_log_format(ab, " new=");
1435 		audit_log_untrustedstring(ab, new);
1436 		audit_log_format(ab, " res=%d", !err);
1437 		audit_log_end(ab);
1438 		kfree(old);
1439 		kfree(new);
1440 		break;
1441 	}
1442 	case AUDIT_SIGNAL_INFO:
1443 		len = 0;
1444 		if (audit_sig_sid) {
1445 			err = security_secid_to_secctx(audit_sig_sid, &ctx, &len);
1446 			if (err)
1447 				return err;
1448 		}
1449 		sig_data = kmalloc(sizeof(*sig_data) + len, GFP_KERNEL);
1450 		if (!sig_data) {
1451 			if (audit_sig_sid)
1452 				security_release_secctx(ctx, len);
1453 			return -ENOMEM;
1454 		}
1455 		sig_data->uid = from_kuid(&init_user_ns, audit_sig_uid);
1456 		sig_data->pid = audit_sig_pid;
1457 		if (audit_sig_sid) {
1458 			memcpy(sig_data->ctx, ctx, len);
1459 			security_release_secctx(ctx, len);
1460 		}
1461 		audit_send_reply(skb, seq, AUDIT_SIGNAL_INFO, 0, 0,
1462 				 sig_data, sizeof(*sig_data) + len);
1463 		kfree(sig_data);
1464 		break;
1465 	case AUDIT_TTY_GET: {
1466 		struct audit_tty_status s;
1467 		unsigned int t;
1468 
1469 		t = READ_ONCE(current->signal->audit_tty);
1470 		s.enabled = t & AUDIT_TTY_ENABLE;
1471 		s.log_passwd = !!(t & AUDIT_TTY_LOG_PASSWD);
1472 
1473 		audit_send_reply(skb, seq, AUDIT_TTY_GET, 0, 0, &s, sizeof(s));
1474 		break;
1475 	}
1476 	case AUDIT_TTY_SET: {
1477 		struct audit_tty_status s, old;
1478 		struct audit_buffer	*ab;
1479 		unsigned int t;
1480 
1481 		memset(&s, 0, sizeof(s));
1482 		/* guard against past and future API changes */
1483 		memcpy(&s, data, min_t(size_t, sizeof(s), data_len));
1484 		/* check if new data is valid */
1485 		if ((s.enabled != 0 && s.enabled != 1) ||
1486 		    (s.log_passwd != 0 && s.log_passwd != 1))
1487 			err = -EINVAL;
1488 
1489 		if (err)
1490 			t = READ_ONCE(current->signal->audit_tty);
1491 		else {
1492 			t = s.enabled | (-s.log_passwd & AUDIT_TTY_LOG_PASSWD);
1493 			t = xchg(&current->signal->audit_tty, t);
1494 		}
1495 		old.enabled = t & AUDIT_TTY_ENABLE;
1496 		old.log_passwd = !!(t & AUDIT_TTY_LOG_PASSWD);
1497 
1498 		audit_log_common_recv_msg(audit_context(), &ab,
1499 					  AUDIT_CONFIG_CHANGE);
1500 		audit_log_format(ab, " op=tty_set old-enabled=%d new-enabled=%d"
1501 				 " old-log_passwd=%d new-log_passwd=%d res=%d",
1502 				 old.enabled, s.enabled, old.log_passwd,
1503 				 s.log_passwd, !err);
1504 		audit_log_end(ab);
1505 		break;
1506 	}
1507 	default:
1508 		err = -EINVAL;
1509 		break;
1510 	}
1511 
1512 	return err < 0 ? err : 0;
1513 }
1514 
1515 /**
1516  * audit_receive - receive messages from a netlink control socket
1517  * @skb: the message buffer
1518  *
1519  * Parse the provided skb and deal with any messages that may be present,
1520  * malformed skbs are discarded.
1521  */
1522 static void audit_receive(struct sk_buff  *skb)
1523 {
1524 	struct nlmsghdr *nlh;
1525 	/*
1526 	 * len MUST be signed for nlmsg_next to be able to dec it below 0
1527 	 * if the nlmsg_len was not aligned
1528 	 */
1529 	int len;
1530 	int err;
1531 
1532 	nlh = nlmsg_hdr(skb);
1533 	len = skb->len;
1534 
1535 	audit_ctl_lock();
1536 	while (nlmsg_ok(nlh, len)) {
1537 		err = audit_receive_msg(skb, nlh);
1538 		/* if err or if this message says it wants a response */
1539 		if (err || (nlh->nlmsg_flags & NLM_F_ACK))
1540 			netlink_ack(skb, nlh, err, NULL);
1541 
1542 		nlh = nlmsg_next(nlh, &len);
1543 	}
1544 	audit_ctl_unlock();
1545 }
1546 
1547 /* Log information about who is connecting to the audit multicast socket */
1548 static void audit_log_multicast(int group, const char *op, int err)
1549 {
1550 	const struct cred *cred;
1551 	struct tty_struct *tty;
1552 	char comm[sizeof(current->comm)];
1553 	struct audit_buffer *ab;
1554 
1555 	if (!audit_enabled)
1556 		return;
1557 
1558 	ab = audit_log_start(audit_context(), GFP_KERNEL, AUDIT_EVENT_LISTENER);
1559 	if (!ab)
1560 		return;
1561 
1562 	cred = current_cred();
1563 	tty = audit_get_tty();
1564 	audit_log_format(ab, "pid=%u uid=%u auid=%u tty=%s ses=%u",
1565 			 task_pid_nr(current),
1566 			 from_kuid(&init_user_ns, cred->uid),
1567 			 from_kuid(&init_user_ns, audit_get_loginuid(current)),
1568 			 tty ? tty_name(tty) : "(none)",
1569 			 audit_get_sessionid(current));
1570 	audit_put_tty(tty);
1571 	audit_log_task_context(ab); /* subj= */
1572 	audit_log_format(ab, " comm=");
1573 	audit_log_untrustedstring(ab, get_task_comm(comm, current));
1574 	audit_log_d_path_exe(ab, current->mm); /* exe= */
1575 	audit_log_format(ab, " nl-mcgrp=%d op=%s res=%d", group, op, !err);
1576 	audit_log_end(ab);
1577 }
1578 
1579 /* Run custom bind function on netlink socket group connect or bind requests. */
1580 static int audit_multicast_bind(struct net *net, int group)
1581 {
1582 	int err = 0;
1583 
1584 	if (!capable(CAP_AUDIT_READ))
1585 		err = -EPERM;
1586 	audit_log_multicast(group, "connect", err);
1587 	return err;
1588 }
1589 
1590 static void audit_multicast_unbind(struct net *net, int group)
1591 {
1592 	audit_log_multicast(group, "disconnect", 0);
1593 }
1594 
1595 static int __net_init audit_net_init(struct net *net)
1596 {
1597 	struct netlink_kernel_cfg cfg = {
1598 		.input	= audit_receive,
1599 		.bind	= audit_multicast_bind,
1600 		.unbind	= audit_multicast_unbind,
1601 		.flags	= NL_CFG_F_NONROOT_RECV,
1602 		.groups	= AUDIT_NLGRP_MAX,
1603 	};
1604 
1605 	struct audit_net *aunet = net_generic(net, audit_net_id);
1606 
1607 	aunet->sk = netlink_kernel_create(net, NETLINK_AUDIT, &cfg);
1608 	if (aunet->sk == NULL) {
1609 		audit_panic("cannot initialize netlink socket in namespace");
1610 		return -ENOMEM;
1611 	}
1612 	aunet->sk->sk_sndtimeo = MAX_SCHEDULE_TIMEOUT;
1613 
1614 	return 0;
1615 }
1616 
1617 static void __net_exit audit_net_exit(struct net *net)
1618 {
1619 	struct audit_net *aunet = net_generic(net, audit_net_id);
1620 
1621 	/* NOTE: you would think that we would want to check the auditd
1622 	 * connection and potentially reset it here if it lives in this
1623 	 * namespace, but since the auditd connection tracking struct holds a
1624 	 * reference to this namespace (see auditd_set()) we are only ever
1625 	 * going to get here after that connection has been released */
1626 
1627 	netlink_kernel_release(aunet->sk);
1628 }
1629 
1630 static struct pernet_operations audit_net_ops __net_initdata = {
1631 	.init = audit_net_init,
1632 	.exit = audit_net_exit,
1633 	.id = &audit_net_id,
1634 	.size = sizeof(struct audit_net),
1635 };
1636 
1637 /* Initialize audit support at boot time. */
1638 static int __init audit_init(void)
1639 {
1640 	int i;
1641 
1642 	if (audit_initialized == AUDIT_DISABLED)
1643 		return 0;
1644 
1645 	audit_buffer_cache = kmem_cache_create("audit_buffer",
1646 					       sizeof(struct audit_buffer),
1647 					       0, SLAB_PANIC, NULL);
1648 
1649 	skb_queue_head_init(&audit_queue);
1650 	skb_queue_head_init(&audit_retry_queue);
1651 	skb_queue_head_init(&audit_hold_queue);
1652 
1653 	for (i = 0; i < AUDIT_INODE_BUCKETS; i++)
1654 		INIT_LIST_HEAD(&audit_inode_hash[i]);
1655 
1656 	mutex_init(&audit_cmd_mutex.lock);
1657 	audit_cmd_mutex.owner = NULL;
1658 
1659 	pr_info("initializing netlink subsys (%s)\n",
1660 		audit_default ? "enabled" : "disabled");
1661 	register_pernet_subsys(&audit_net_ops);
1662 
1663 	audit_initialized = AUDIT_INITIALIZED;
1664 
1665 	kauditd_task = kthread_run(kauditd_thread, NULL, "kauditd");
1666 	if (IS_ERR(kauditd_task)) {
1667 		int err = PTR_ERR(kauditd_task);
1668 		panic("audit: failed to start the kauditd thread (%d)\n", err);
1669 	}
1670 
1671 	audit_log(NULL, GFP_KERNEL, AUDIT_KERNEL,
1672 		"state=initialized audit_enabled=%u res=1",
1673 		 audit_enabled);
1674 
1675 	return 0;
1676 }
1677 postcore_initcall(audit_init);
1678 
1679 /*
1680  * Process kernel command-line parameter at boot time.
1681  * audit={0|off} or audit={1|on}.
1682  */
1683 static int __init audit_enable(char *str)
1684 {
1685 	if (!strcasecmp(str, "off") || !strcmp(str, "0"))
1686 		audit_default = AUDIT_OFF;
1687 	else if (!strcasecmp(str, "on") || !strcmp(str, "1"))
1688 		audit_default = AUDIT_ON;
1689 	else {
1690 		pr_err("audit: invalid 'audit' parameter value (%s)\n", str);
1691 		audit_default = AUDIT_ON;
1692 	}
1693 
1694 	if (audit_default == AUDIT_OFF)
1695 		audit_initialized = AUDIT_DISABLED;
1696 	if (audit_set_enabled(audit_default))
1697 		pr_err("audit: error setting audit state (%d)\n",
1698 		       audit_default);
1699 
1700 	pr_info("%s\n", audit_default ?
1701 		"enabled (after initialization)" : "disabled (until reboot)");
1702 
1703 	return 1;
1704 }
1705 __setup("audit=", audit_enable);
1706 
1707 /* Process kernel command-line parameter at boot time.
1708  * audit_backlog_limit=<n> */
1709 static int __init audit_backlog_limit_set(char *str)
1710 {
1711 	u32 audit_backlog_limit_arg;
1712 
1713 	pr_info("audit_backlog_limit: ");
1714 	if (kstrtouint(str, 0, &audit_backlog_limit_arg)) {
1715 		pr_cont("using default of %u, unable to parse %s\n",
1716 			audit_backlog_limit, str);
1717 		return 1;
1718 	}
1719 
1720 	audit_backlog_limit = audit_backlog_limit_arg;
1721 	pr_cont("%d\n", audit_backlog_limit);
1722 
1723 	return 1;
1724 }
1725 __setup("audit_backlog_limit=", audit_backlog_limit_set);
1726 
1727 static void audit_buffer_free(struct audit_buffer *ab)
1728 {
1729 	if (!ab)
1730 		return;
1731 
1732 	kfree_skb(ab->skb);
1733 	kmem_cache_free(audit_buffer_cache, ab);
1734 }
1735 
1736 static struct audit_buffer *audit_buffer_alloc(struct audit_context *ctx,
1737 					       gfp_t gfp_mask, int type)
1738 {
1739 	struct audit_buffer *ab;
1740 
1741 	ab = kmem_cache_alloc(audit_buffer_cache, gfp_mask);
1742 	if (!ab)
1743 		return NULL;
1744 
1745 	ab->skb = nlmsg_new(AUDIT_BUFSIZ, gfp_mask);
1746 	if (!ab->skb)
1747 		goto err;
1748 	if (!nlmsg_put(ab->skb, 0, 0, type, 0, 0))
1749 		goto err;
1750 
1751 	ab->ctx = ctx;
1752 	ab->gfp_mask = gfp_mask;
1753 
1754 	return ab;
1755 
1756 err:
1757 	audit_buffer_free(ab);
1758 	return NULL;
1759 }
1760 
1761 /**
1762  * audit_serial - compute a serial number for the audit record
1763  *
1764  * Compute a serial number for the audit record.  Audit records are
1765  * written to user-space as soon as they are generated, so a complete
1766  * audit record may be written in several pieces.  The timestamp of the
1767  * record and this serial number are used by the user-space tools to
1768  * determine which pieces belong to the same audit record.  The
1769  * (timestamp,serial) tuple is unique for each syscall and is live from
1770  * syscall entry to syscall exit.
1771  *
1772  * NOTE: Another possibility is to store the formatted records off the
1773  * audit context (for those records that have a context), and emit them
1774  * all at syscall exit.  However, this could delay the reporting of
1775  * significant errors until syscall exit (or never, if the system
1776  * halts).
1777  */
1778 unsigned int audit_serial(void)
1779 {
1780 	static atomic_t serial = ATOMIC_INIT(0);
1781 
1782 	return atomic_inc_return(&serial);
1783 }
1784 
1785 static inline void audit_get_stamp(struct audit_context *ctx,
1786 				   struct timespec64 *t, unsigned int *serial)
1787 {
1788 	if (!ctx || !auditsc_get_stamp(ctx, t, serial)) {
1789 		ktime_get_coarse_real_ts64(t);
1790 		*serial = audit_serial();
1791 	}
1792 }
1793 
1794 /**
1795  * audit_log_start - obtain an audit buffer
1796  * @ctx: audit_context (may be NULL)
1797  * @gfp_mask: type of allocation
1798  * @type: audit message type
1799  *
1800  * Returns audit_buffer pointer on success or NULL on error.
1801  *
1802  * Obtain an audit buffer.  This routine does locking to obtain the
1803  * audit buffer, but then no locking is required for calls to
1804  * audit_log_*format.  If the task (ctx) is a task that is currently in a
1805  * syscall, then the syscall is marked as auditable and an audit record
1806  * will be written at syscall exit.  If there is no associated task, then
1807  * task context (ctx) should be NULL.
1808  */
1809 struct audit_buffer *audit_log_start(struct audit_context *ctx, gfp_t gfp_mask,
1810 				     int type)
1811 {
1812 	struct audit_buffer *ab;
1813 	struct timespec64 t;
1814 	unsigned int serial;
1815 
1816 	if (audit_initialized != AUDIT_INITIALIZED)
1817 		return NULL;
1818 
1819 	if (unlikely(!audit_filter(type, AUDIT_FILTER_EXCLUDE)))
1820 		return NULL;
1821 
1822 	/* NOTE: don't ever fail/sleep on these two conditions:
1823 	 * 1. auditd generated record - since we need auditd to drain the
1824 	 *    queue; also, when we are checking for auditd, compare PIDs using
1825 	 *    task_tgid_vnr() since auditd_pid is set in audit_receive_msg()
1826 	 *    using a PID anchored in the caller's namespace
1827 	 * 2. generator holding the audit_cmd_mutex - we don't want to block
1828 	 *    while holding the mutex */
1829 	if (!(auditd_test_task(current) || audit_ctl_owner_current())) {
1830 		long stime = audit_backlog_wait_time;
1831 
1832 		while (audit_backlog_limit &&
1833 		       (skb_queue_len(&audit_queue) > audit_backlog_limit)) {
1834 			/* wake kauditd to try and flush the queue */
1835 			wake_up_interruptible(&kauditd_wait);
1836 
1837 			/* sleep if we are allowed and we haven't exhausted our
1838 			 * backlog wait limit */
1839 			if (gfpflags_allow_blocking(gfp_mask) && (stime > 0)) {
1840 				long rtime = stime;
1841 
1842 				DECLARE_WAITQUEUE(wait, current);
1843 
1844 				add_wait_queue_exclusive(&audit_backlog_wait,
1845 							 &wait);
1846 				set_current_state(TASK_UNINTERRUPTIBLE);
1847 				stime = schedule_timeout(rtime);
1848 				atomic_add(rtime - stime, &audit_backlog_wait_time_actual);
1849 				remove_wait_queue(&audit_backlog_wait, &wait);
1850 			} else {
1851 				if (audit_rate_check() && printk_ratelimit())
1852 					pr_warn("audit_backlog=%d > audit_backlog_limit=%d\n",
1853 						skb_queue_len(&audit_queue),
1854 						audit_backlog_limit);
1855 				audit_log_lost("backlog limit exceeded");
1856 				return NULL;
1857 			}
1858 		}
1859 	}
1860 
1861 	ab = audit_buffer_alloc(ctx, gfp_mask, type);
1862 	if (!ab) {
1863 		audit_log_lost("out of memory in audit_log_start");
1864 		return NULL;
1865 	}
1866 
1867 	audit_get_stamp(ab->ctx, &t, &serial);
1868 	/* cancel dummy context to enable supporting records */
1869 	if (ctx)
1870 		ctx->dummy = 0;
1871 	audit_log_format(ab, "audit(%llu.%03lu:%u): ",
1872 			 (unsigned long long)t.tv_sec, t.tv_nsec/1000000, serial);
1873 
1874 	return ab;
1875 }
1876 
1877 /**
1878  * audit_expand - expand skb in the audit buffer
1879  * @ab: audit_buffer
1880  * @extra: space to add at tail of the skb
1881  *
1882  * Returns 0 (no space) on failed expansion, or available space if
1883  * successful.
1884  */
1885 static inline int audit_expand(struct audit_buffer *ab, int extra)
1886 {
1887 	struct sk_buff *skb = ab->skb;
1888 	int oldtail = skb_tailroom(skb);
1889 	int ret = pskb_expand_head(skb, 0, extra, ab->gfp_mask);
1890 	int newtail = skb_tailroom(skb);
1891 
1892 	if (ret < 0) {
1893 		audit_log_lost("out of memory in audit_expand");
1894 		return 0;
1895 	}
1896 
1897 	skb->truesize += newtail - oldtail;
1898 	return newtail;
1899 }
1900 
1901 /*
1902  * Format an audit message into the audit buffer.  If there isn't enough
1903  * room in the audit buffer, more room will be allocated and vsnprint
1904  * will be called a second time.  Currently, we assume that a printk
1905  * can't format message larger than 1024 bytes, so we don't either.
1906  */
1907 static void audit_log_vformat(struct audit_buffer *ab, const char *fmt,
1908 			      va_list args)
1909 {
1910 	int len, avail;
1911 	struct sk_buff *skb;
1912 	va_list args2;
1913 
1914 	if (!ab)
1915 		return;
1916 
1917 	BUG_ON(!ab->skb);
1918 	skb = ab->skb;
1919 	avail = skb_tailroom(skb);
1920 	if (avail == 0) {
1921 		avail = audit_expand(ab, AUDIT_BUFSIZ);
1922 		if (!avail)
1923 			goto out;
1924 	}
1925 	va_copy(args2, args);
1926 	len = vsnprintf(skb_tail_pointer(skb), avail, fmt, args);
1927 	if (len >= avail) {
1928 		/* The printk buffer is 1024 bytes long, so if we get
1929 		 * here and AUDIT_BUFSIZ is at least 1024, then we can
1930 		 * log everything that printk could have logged. */
1931 		avail = audit_expand(ab,
1932 			max_t(unsigned, AUDIT_BUFSIZ, 1+len-avail));
1933 		if (!avail)
1934 			goto out_va_end;
1935 		len = vsnprintf(skb_tail_pointer(skb), avail, fmt, args2);
1936 	}
1937 	if (len > 0)
1938 		skb_put(skb, len);
1939 out_va_end:
1940 	va_end(args2);
1941 out:
1942 	return;
1943 }
1944 
1945 /**
1946  * audit_log_format - format a message into the audit buffer.
1947  * @ab: audit_buffer
1948  * @fmt: format string
1949  * @...: optional parameters matching @fmt string
1950  *
1951  * All the work is done in audit_log_vformat.
1952  */
1953 void audit_log_format(struct audit_buffer *ab, const char *fmt, ...)
1954 {
1955 	va_list args;
1956 
1957 	if (!ab)
1958 		return;
1959 	va_start(args, fmt);
1960 	audit_log_vformat(ab, fmt, args);
1961 	va_end(args);
1962 }
1963 
1964 /**
1965  * audit_log_n_hex - convert a buffer to hex and append it to the audit skb
1966  * @ab: the audit_buffer
1967  * @buf: buffer to convert to hex
1968  * @len: length of @buf to be converted
1969  *
1970  * No return value; failure to expand is silently ignored.
1971  *
1972  * This function will take the passed buf and convert it into a string of
1973  * ascii hex digits. The new string is placed onto the skb.
1974  */
1975 void audit_log_n_hex(struct audit_buffer *ab, const unsigned char *buf,
1976 		size_t len)
1977 {
1978 	int i, avail, new_len;
1979 	unsigned char *ptr;
1980 	struct sk_buff *skb;
1981 
1982 	if (!ab)
1983 		return;
1984 
1985 	BUG_ON(!ab->skb);
1986 	skb = ab->skb;
1987 	avail = skb_tailroom(skb);
1988 	new_len = len<<1;
1989 	if (new_len >= avail) {
1990 		/* Round the buffer request up to the next multiple */
1991 		new_len = AUDIT_BUFSIZ*(((new_len-avail)/AUDIT_BUFSIZ) + 1);
1992 		avail = audit_expand(ab, new_len);
1993 		if (!avail)
1994 			return;
1995 	}
1996 
1997 	ptr = skb_tail_pointer(skb);
1998 	for (i = 0; i < len; i++)
1999 		ptr = hex_byte_pack_upper(ptr, buf[i]);
2000 	*ptr = 0;
2001 	skb_put(skb, len << 1); /* new string is twice the old string */
2002 }
2003 
2004 /*
2005  * Format a string of no more than slen characters into the audit buffer,
2006  * enclosed in quote marks.
2007  */
2008 void audit_log_n_string(struct audit_buffer *ab, const char *string,
2009 			size_t slen)
2010 {
2011 	int avail, new_len;
2012 	unsigned char *ptr;
2013 	struct sk_buff *skb;
2014 
2015 	if (!ab)
2016 		return;
2017 
2018 	BUG_ON(!ab->skb);
2019 	skb = ab->skb;
2020 	avail = skb_tailroom(skb);
2021 	new_len = slen + 3;	/* enclosing quotes + null terminator */
2022 	if (new_len > avail) {
2023 		avail = audit_expand(ab, new_len);
2024 		if (!avail)
2025 			return;
2026 	}
2027 	ptr = skb_tail_pointer(skb);
2028 	*ptr++ = '"';
2029 	memcpy(ptr, string, slen);
2030 	ptr += slen;
2031 	*ptr++ = '"';
2032 	*ptr = 0;
2033 	skb_put(skb, slen + 2);	/* don't include null terminator */
2034 }
2035 
2036 /**
2037  * audit_string_contains_control - does a string need to be logged in hex
2038  * @string: string to be checked
2039  * @len: max length of the string to check
2040  */
2041 bool audit_string_contains_control(const char *string, size_t len)
2042 {
2043 	const unsigned char *p;
2044 	for (p = string; p < (const unsigned char *)string + len; p++) {
2045 		if (*p == '"' || *p < 0x21 || *p > 0x7e)
2046 			return true;
2047 	}
2048 	return false;
2049 }
2050 
2051 /**
2052  * audit_log_n_untrustedstring - log a string that may contain random characters
2053  * @ab: audit_buffer
2054  * @len: length of string (not including trailing null)
2055  * @string: string to be logged
2056  *
2057  * This code will escape a string that is passed to it if the string
2058  * contains a control character, unprintable character, double quote mark,
2059  * or a space. Unescaped strings will start and end with a double quote mark.
2060  * Strings that are escaped are printed in hex (2 digits per char).
2061  *
2062  * The caller specifies the number of characters in the string to log, which may
2063  * or may not be the entire string.
2064  */
2065 void audit_log_n_untrustedstring(struct audit_buffer *ab, const char *string,
2066 				 size_t len)
2067 {
2068 	if (audit_string_contains_control(string, len))
2069 		audit_log_n_hex(ab, string, len);
2070 	else
2071 		audit_log_n_string(ab, string, len);
2072 }
2073 
2074 /**
2075  * audit_log_untrustedstring - log a string that may contain random characters
2076  * @ab: audit_buffer
2077  * @string: string to be logged
2078  *
2079  * Same as audit_log_n_untrustedstring(), except that strlen is used to
2080  * determine string length.
2081  */
2082 void audit_log_untrustedstring(struct audit_buffer *ab, const char *string)
2083 {
2084 	audit_log_n_untrustedstring(ab, string, strlen(string));
2085 }
2086 
2087 /* This is a helper-function to print the escaped d_path */
2088 void audit_log_d_path(struct audit_buffer *ab, const char *prefix,
2089 		      const struct path *path)
2090 {
2091 	char *p, *pathname;
2092 
2093 	if (prefix)
2094 		audit_log_format(ab, "%s", prefix);
2095 
2096 	/* We will allow 11 spaces for ' (deleted)' to be appended */
2097 	pathname = kmalloc(PATH_MAX+11, ab->gfp_mask);
2098 	if (!pathname) {
2099 		audit_log_format(ab, "\"<no_memory>\"");
2100 		return;
2101 	}
2102 	p = d_path(path, pathname, PATH_MAX+11);
2103 	if (IS_ERR(p)) { /* Should never happen since we send PATH_MAX */
2104 		/* FIXME: can we save some information here? */
2105 		audit_log_format(ab, "\"<too_long>\"");
2106 	} else
2107 		audit_log_untrustedstring(ab, p);
2108 	kfree(pathname);
2109 }
2110 
2111 void audit_log_session_info(struct audit_buffer *ab)
2112 {
2113 	unsigned int sessionid = audit_get_sessionid(current);
2114 	uid_t auid = from_kuid(&init_user_ns, audit_get_loginuid(current));
2115 
2116 	audit_log_format(ab, "auid=%u ses=%u", auid, sessionid);
2117 }
2118 
2119 void audit_log_key(struct audit_buffer *ab, char *key)
2120 {
2121 	audit_log_format(ab, " key=");
2122 	if (key)
2123 		audit_log_untrustedstring(ab, key);
2124 	else
2125 		audit_log_format(ab, "(null)");
2126 }
2127 
2128 int audit_log_task_context(struct audit_buffer *ab)
2129 {
2130 	char *ctx = NULL;
2131 	unsigned len;
2132 	int error;
2133 	u32 sid;
2134 
2135 	security_task_getsecid_subj(current, &sid);
2136 	if (!sid)
2137 		return 0;
2138 
2139 	error = security_secid_to_secctx(sid, &ctx, &len);
2140 	if (error) {
2141 		if (error != -EINVAL)
2142 			goto error_path;
2143 		return 0;
2144 	}
2145 
2146 	audit_log_format(ab, " subj=%s", ctx);
2147 	security_release_secctx(ctx, len);
2148 	return 0;
2149 
2150 error_path:
2151 	audit_panic("error in audit_log_task_context");
2152 	return error;
2153 }
2154 EXPORT_SYMBOL(audit_log_task_context);
2155 
2156 void audit_log_d_path_exe(struct audit_buffer *ab,
2157 			  struct mm_struct *mm)
2158 {
2159 	struct file *exe_file;
2160 
2161 	if (!mm)
2162 		goto out_null;
2163 
2164 	exe_file = get_mm_exe_file(mm);
2165 	if (!exe_file)
2166 		goto out_null;
2167 
2168 	audit_log_d_path(ab, " exe=", &exe_file->f_path);
2169 	fput(exe_file);
2170 	return;
2171 out_null:
2172 	audit_log_format(ab, " exe=(null)");
2173 }
2174 
2175 struct tty_struct *audit_get_tty(void)
2176 {
2177 	struct tty_struct *tty = NULL;
2178 	unsigned long flags;
2179 
2180 	spin_lock_irqsave(&current->sighand->siglock, flags);
2181 	if (current->signal)
2182 		tty = tty_kref_get(current->signal->tty);
2183 	spin_unlock_irqrestore(&current->sighand->siglock, flags);
2184 	return tty;
2185 }
2186 
2187 void audit_put_tty(struct tty_struct *tty)
2188 {
2189 	tty_kref_put(tty);
2190 }
2191 
2192 void audit_log_task_info(struct audit_buffer *ab)
2193 {
2194 	const struct cred *cred;
2195 	char comm[sizeof(current->comm)];
2196 	struct tty_struct *tty;
2197 
2198 	if (!ab)
2199 		return;
2200 
2201 	cred = current_cred();
2202 	tty = audit_get_tty();
2203 	audit_log_format(ab,
2204 			 " ppid=%d pid=%d auid=%u uid=%u gid=%u"
2205 			 " euid=%u suid=%u fsuid=%u"
2206 			 " egid=%u sgid=%u fsgid=%u tty=%s ses=%u",
2207 			 task_ppid_nr(current),
2208 			 task_tgid_nr(current),
2209 			 from_kuid(&init_user_ns, audit_get_loginuid(current)),
2210 			 from_kuid(&init_user_ns, cred->uid),
2211 			 from_kgid(&init_user_ns, cred->gid),
2212 			 from_kuid(&init_user_ns, cred->euid),
2213 			 from_kuid(&init_user_ns, cred->suid),
2214 			 from_kuid(&init_user_ns, cred->fsuid),
2215 			 from_kgid(&init_user_ns, cred->egid),
2216 			 from_kgid(&init_user_ns, cred->sgid),
2217 			 from_kgid(&init_user_ns, cred->fsgid),
2218 			 tty ? tty_name(tty) : "(none)",
2219 			 audit_get_sessionid(current));
2220 	audit_put_tty(tty);
2221 	audit_log_format(ab, " comm=");
2222 	audit_log_untrustedstring(ab, get_task_comm(comm, current));
2223 	audit_log_d_path_exe(ab, current->mm);
2224 	audit_log_task_context(ab);
2225 }
2226 EXPORT_SYMBOL(audit_log_task_info);
2227 
2228 /**
2229  * audit_log_path_denied - report a path restriction denial
2230  * @type: audit message type (AUDIT_ANOM_LINK, AUDIT_ANOM_CREAT, etc)
2231  * @operation: specific operation name
2232  */
2233 void audit_log_path_denied(int type, const char *operation)
2234 {
2235 	struct audit_buffer *ab;
2236 
2237 	if (!audit_enabled || audit_dummy_context())
2238 		return;
2239 
2240 	/* Generate log with subject, operation, outcome. */
2241 	ab = audit_log_start(audit_context(), GFP_KERNEL, type);
2242 	if (!ab)
2243 		return;
2244 	audit_log_format(ab, "op=%s", operation);
2245 	audit_log_task_info(ab);
2246 	audit_log_format(ab, " res=0");
2247 	audit_log_end(ab);
2248 }
2249 
2250 /* global counter which is incremented every time something logs in */
2251 static atomic_t session_id = ATOMIC_INIT(0);
2252 
2253 static int audit_set_loginuid_perm(kuid_t loginuid)
2254 {
2255 	/* if we are unset, we don't need privs */
2256 	if (!audit_loginuid_set(current))
2257 		return 0;
2258 	/* if AUDIT_FEATURE_LOGINUID_IMMUTABLE means never ever allow a change*/
2259 	if (is_audit_feature_set(AUDIT_FEATURE_LOGINUID_IMMUTABLE))
2260 		return -EPERM;
2261 	/* it is set, you need permission */
2262 	if (!capable(CAP_AUDIT_CONTROL))
2263 		return -EPERM;
2264 	/* reject if this is not an unset and we don't allow that */
2265 	if (is_audit_feature_set(AUDIT_FEATURE_ONLY_UNSET_LOGINUID)
2266 				 && uid_valid(loginuid))
2267 		return -EPERM;
2268 	return 0;
2269 }
2270 
2271 static void audit_log_set_loginuid(kuid_t koldloginuid, kuid_t kloginuid,
2272 				   unsigned int oldsessionid,
2273 				   unsigned int sessionid, int rc)
2274 {
2275 	struct audit_buffer *ab;
2276 	uid_t uid, oldloginuid, loginuid;
2277 	struct tty_struct *tty;
2278 
2279 	if (!audit_enabled)
2280 		return;
2281 
2282 	ab = audit_log_start(audit_context(), GFP_KERNEL, AUDIT_LOGIN);
2283 	if (!ab)
2284 		return;
2285 
2286 	uid = from_kuid(&init_user_ns, task_uid(current));
2287 	oldloginuid = from_kuid(&init_user_ns, koldloginuid);
2288 	loginuid = from_kuid(&init_user_ns, kloginuid);
2289 	tty = audit_get_tty();
2290 
2291 	audit_log_format(ab, "pid=%d uid=%u", task_tgid_nr(current), uid);
2292 	audit_log_task_context(ab);
2293 	audit_log_format(ab, " old-auid=%u auid=%u tty=%s old-ses=%u ses=%u res=%d",
2294 			 oldloginuid, loginuid, tty ? tty_name(tty) : "(none)",
2295 			 oldsessionid, sessionid, !rc);
2296 	audit_put_tty(tty);
2297 	audit_log_end(ab);
2298 }
2299 
2300 /**
2301  * audit_set_loginuid - set current task's loginuid
2302  * @loginuid: loginuid value
2303  *
2304  * Returns 0.
2305  *
2306  * Called (set) from fs/proc/base.c::proc_loginuid_write().
2307  */
2308 int audit_set_loginuid(kuid_t loginuid)
2309 {
2310 	unsigned int oldsessionid, sessionid = AUDIT_SID_UNSET;
2311 	kuid_t oldloginuid;
2312 	int rc;
2313 
2314 	oldloginuid = audit_get_loginuid(current);
2315 	oldsessionid = audit_get_sessionid(current);
2316 
2317 	rc = audit_set_loginuid_perm(loginuid);
2318 	if (rc)
2319 		goto out;
2320 
2321 	/* are we setting or clearing? */
2322 	if (uid_valid(loginuid)) {
2323 		sessionid = (unsigned int)atomic_inc_return(&session_id);
2324 		if (unlikely(sessionid == AUDIT_SID_UNSET))
2325 			sessionid = (unsigned int)atomic_inc_return(&session_id);
2326 	}
2327 
2328 	current->sessionid = sessionid;
2329 	current->loginuid = loginuid;
2330 out:
2331 	audit_log_set_loginuid(oldloginuid, loginuid, oldsessionid, sessionid, rc);
2332 	return rc;
2333 }
2334 
2335 /**
2336  * audit_signal_info - record signal info for shutting down audit subsystem
2337  * @sig: signal value
2338  * @t: task being signaled
2339  *
2340  * If the audit subsystem is being terminated, record the task (pid)
2341  * and uid that is doing that.
2342  */
2343 int audit_signal_info(int sig, struct task_struct *t)
2344 {
2345 	kuid_t uid = current_uid(), auid;
2346 
2347 	if (auditd_test_task(t) &&
2348 	    (sig == SIGTERM || sig == SIGHUP ||
2349 	     sig == SIGUSR1 || sig == SIGUSR2)) {
2350 		audit_sig_pid = task_tgid_nr(current);
2351 		auid = audit_get_loginuid(current);
2352 		if (uid_valid(auid))
2353 			audit_sig_uid = auid;
2354 		else
2355 			audit_sig_uid = uid;
2356 		security_task_getsecid_subj(current, &audit_sig_sid);
2357 	}
2358 
2359 	return audit_signal_info_syscall(t);
2360 }
2361 
2362 /**
2363  * audit_log_end - end one audit record
2364  * @ab: the audit_buffer
2365  *
2366  * We can not do a netlink send inside an irq context because it blocks (last
2367  * arg, flags, is not set to MSG_DONTWAIT), so the audit buffer is placed on a
2368  * queue and a kthread is scheduled to remove them from the queue outside the
2369  * irq context.  May be called in any context.
2370  */
2371 void audit_log_end(struct audit_buffer *ab)
2372 {
2373 	struct sk_buff *skb;
2374 	struct nlmsghdr *nlh;
2375 
2376 	if (!ab)
2377 		return;
2378 
2379 	if (audit_rate_check()) {
2380 		skb = ab->skb;
2381 		ab->skb = NULL;
2382 
2383 		/* setup the netlink header, see the comments in
2384 		 * kauditd_send_multicast_skb() for length quirks */
2385 		nlh = nlmsg_hdr(skb);
2386 		nlh->nlmsg_len = skb->len - NLMSG_HDRLEN;
2387 
2388 		/* queue the netlink packet and poke the kauditd thread */
2389 		skb_queue_tail(&audit_queue, skb);
2390 		wake_up_interruptible(&kauditd_wait);
2391 	} else
2392 		audit_log_lost("rate limit exceeded");
2393 
2394 	audit_buffer_free(ab);
2395 }
2396 
2397 /**
2398  * audit_log - Log an audit record
2399  * @ctx: audit context
2400  * @gfp_mask: type of allocation
2401  * @type: audit message type
2402  * @fmt: format string to use
2403  * @...: variable parameters matching the format string
2404  *
2405  * This is a convenience function that calls audit_log_start,
2406  * audit_log_vformat, and audit_log_end.  It may be called
2407  * in any context.
2408  */
2409 void audit_log(struct audit_context *ctx, gfp_t gfp_mask, int type,
2410 	       const char *fmt, ...)
2411 {
2412 	struct audit_buffer *ab;
2413 	va_list args;
2414 
2415 	ab = audit_log_start(ctx, gfp_mask, type);
2416 	if (ab) {
2417 		va_start(args, fmt);
2418 		audit_log_vformat(ab, fmt, args);
2419 		va_end(args);
2420 		audit_log_end(ab);
2421 	}
2422 }
2423 
2424 EXPORT_SYMBOL(audit_log_start);
2425 EXPORT_SYMBOL(audit_log_end);
2426 EXPORT_SYMBOL(audit_log_format);
2427 EXPORT_SYMBOL(audit_log);
2428